Home | History | Annotate | Line # | Download | only in kern
kern_ktrace.c revision 1.72
      1 /*	$NetBSD: kern_ktrace.c,v 1.72 2003/06/28 14:21:54 darrenr Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1989, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)kern_ktrace.c	8.5 (Berkeley) 5/14/95
     36  */
     37 
     38 #include <sys/cdefs.h>
     39 __KERNEL_RCSID(0, "$NetBSD: kern_ktrace.c,v 1.72 2003/06/28 14:21:54 darrenr Exp $");
     40 
     41 #include "opt_ktrace.h"
     42 #include "opt_compat_mach.h"
     43 
     44 #include <sys/param.h>
     45 #include <sys/systm.h>
     46 #include <sys/proc.h>
     47 #include <sys/file.h>
     48 #include <sys/namei.h>
     49 #include <sys/vnode.h>
     50 #include <sys/ktrace.h>
     51 #include <sys/malloc.h>
     52 #include <sys/syslog.h>
     53 #include <sys/filedesc.h>
     54 #include <sys/ioctl.h>
     55 
     56 #include <sys/mount.h>
     57 #include <sys/sa.h>
     58 #include <sys/syscallargs.h>
     59 
     60 #ifdef KTRACE
     61 
     62 int	ktrace_common(struct lwp *, int, int, int, struct file *);
     63 void	ktrinitheader(struct ktr_header *, struct lwp *, int);
     64 int	ktrops(struct lwp *, struct proc *, int, int, struct file *);
     65 int	ktrsetchildren(struct lwp *, struct proc *, int, int,
     66 	    struct file *);
     67 int	ktrwrite(struct lwp *, struct ktr_header *);
     68 int	ktrcanset(struct proc *, struct proc *);
     69 int	ktrsamefile(struct file *, struct file *);
     70 
     71 /*
     72  * "deep" compare of two files for the purposes of clearing a trace.
     73  * Returns true if they're the same open file, or if they point at the
     74  * same underlying vnode/socket.
     75  */
     76 
     77 int
     78 ktrsamefile(f1, f2)
     79 	struct file *f1;
     80 	struct file *f2;
     81 {
     82 	return ((f1 == f2) ||
     83 	    ((f1 != NULL) && (f2 != NULL) &&
     84 		(f1->f_type == f2->f_type) &&
     85 		(f1->f_data == f2->f_data)));
     86 }
     87 
     88 void
     89 ktrderef(p)
     90 	struct proc *p;
     91 {
     92 	struct file *fp = p->p_tracep;
     93 	p->p_traceflag = 0;
     94 	if (fp == NULL)
     95 		return;
     96 	simple_lock(&fp->f_slock);
     97 	FILE_USE(fp);
     98 
     99 	/*
    100 	 * ktrace file descriptor can't be watched (are not visible to
    101 	 * userspace), so no kqueue stuff here
    102 	 */
    103 	closef(fp, NULL);
    104 
    105 	p->p_tracep = NULL;
    106 }
    107 
    108 void
    109 ktradref(p)
    110 	struct proc *p;
    111 {
    112 	struct file *fp = p->p_tracep;
    113 
    114 	fp->f_count++;
    115 }
    116 
    117 void
    118 ktrinitheader(kth, l, type)
    119 	struct ktr_header *kth;
    120 	struct lwp *l;
    121 	int type;
    122 {
    123 	struct proc *p = l->l_proc;
    124 
    125 	memset(kth, 0, sizeof(*kth));
    126 	kth->ktr_type = type;
    127 	microtime(&kth->ktr_time);
    128 	kth->ktr_pid = p->p_pid;
    129 	memcpy(kth->ktr_comm, p->p_comm, MAXCOMLEN);
    130 }
    131 
    132 void
    133 ktrsyscall(l, code, realcode, callp, args)
    134 	struct lwp *l;
    135 	register_t code;
    136 	register_t realcode;
    137 	const struct sysent *callp;
    138 	register_t args[];
    139 {
    140 	struct proc *p = l->l_proc;
    141 	struct ktr_syscall *ktp;
    142 	struct ktr_header kth;
    143 	register_t *argp;
    144 	int argsize;
    145 	size_t len;
    146 	u_int i;
    147 
    148 	if (callp == NULL)
    149 		callp = p->p_emul->e_sysent;
    150 
    151 	argsize = callp[code].sy_narg * sizeof (register_t);
    152 	len = sizeof(struct ktr_syscall) + argsize;
    153 
    154 	p->p_traceflag |= KTRFAC_ACTIVE;
    155 	ktrinitheader(&kth, l, KTR_SYSCALL);
    156 	ktp = malloc(len, M_TEMP, M_WAITOK);
    157 	ktp->ktr_code = realcode;
    158 	ktp->ktr_argsize = argsize;
    159 	argp = (register_t *)((char *)ktp + sizeof(struct ktr_syscall));
    160 	for (i = 0; i < (argsize / sizeof(*argp)); i++)
    161 		*argp++ = args[i];
    162 	kth.ktr_buf = (caddr_t)ktp;
    163 	kth.ktr_len = len;
    164 	(void) ktrwrite(l, &kth);
    165 	free(ktp, M_TEMP);
    166 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    167 }
    168 
    169 void
    170 ktrsysret(l, code, error, retval)
    171 	struct lwp *l;
    172 	register_t code;
    173 	int error;
    174 	register_t *retval;
    175 {
    176 	struct proc *p = l->l_proc;
    177 	struct ktr_header kth;
    178 	struct ktr_sysret ktp;
    179 
    180 	p->p_traceflag |= KTRFAC_ACTIVE;
    181 	ktrinitheader(&kth, l, KTR_SYSRET);
    182 	ktp.ktr_code = code;
    183 	ktp.ktr_eosys = 0;			/* XXX unused */
    184 	ktp.ktr_error = error;
    185 	ktp.ktr_retval = retval ? retval[0] : 0;
    186 	ktp.ktr_retval_1 = retval ? retval[1] : 0;
    187 
    188 	kth.ktr_buf = (caddr_t)&ktp;
    189 	kth.ktr_len = sizeof(struct ktr_sysret);
    190 
    191 	(void) ktrwrite(l, &kth);
    192 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    193 }
    194 
    195 void
    196 ktrnamei(l, path)
    197 	struct lwp *l;
    198 	char *path;
    199 {
    200 	struct proc *p = l->l_proc;
    201 	struct ktr_header kth;
    202 
    203 	p->p_traceflag |= KTRFAC_ACTIVE;
    204 	ktrinitheader(&kth, l, KTR_NAMEI);
    205 	kth.ktr_len = strlen(path);
    206 	kth.ktr_buf = path;
    207 
    208 	(void) ktrwrite(l, &kth);
    209 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    210 }
    211 
    212 void
    213 ktremul(l)
    214 	struct lwp *l;
    215 {
    216 	struct ktr_header kth;
    217 	const char *emul;
    218 	struct proc *p;
    219 
    220 	p = l->l_proc;
    221 	emul = p->p_emul->e_name;
    222 	p->p_traceflag |= KTRFAC_ACTIVE;
    223 	ktrinitheader(&kth, l, KTR_EMUL);
    224 	kth.ktr_len = strlen(emul);
    225 	kth.ktr_buf = (caddr_t)emul;
    226 
    227 	(void) ktrwrite(l, &kth);
    228 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    229 }
    230 
    231 void
    232 ktrgenio(l, fd, rw, iov, len, error)
    233 	struct lwp *l;
    234 	int fd;
    235 	enum uio_rw rw;
    236 	struct iovec *iov;
    237 	int len;
    238 	int error;
    239 {
    240 	struct proc *p = l->l_proc;
    241 	struct ktr_header kth;
    242 	struct ktr_genio *ktp;
    243 	int resid = len, cnt;
    244 	caddr_t cp;
    245 	int buflen;
    246 
    247 	if (error)
    248 		return;
    249 
    250 	p->p_traceflag |= KTRFAC_ACTIVE;
    251 
    252 	buflen = min(PAGE_SIZE, len + sizeof(struct ktr_genio));
    253 
    254 	ktrinitheader(&kth, l, KTR_GENIO);
    255 	ktp = malloc(buflen, M_TEMP, M_WAITOK);
    256 	ktp->ktr_fd = fd;
    257 	ktp->ktr_rw = rw;
    258 
    259 	kth.ktr_buf = (caddr_t)ktp;
    260 
    261 	cp = (caddr_t)((char *)ktp + sizeof(struct ktr_genio));
    262 	buflen -= sizeof(struct ktr_genio);
    263 
    264 	while (resid > 0) {
    265 #if 0 /* XXX NJWLWP */
    266 		KDASSERT(p->p_cpu != NULL);
    267 		KDASSERT(p->p_cpu == curcpu());
    268 #endif
    269 		/* XXX NJWLWP */
    270 		if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD)
    271 			preempt(1);
    272 
    273 		cnt = min(iov->iov_len, buflen);
    274 		if (cnt > resid)
    275 			cnt = resid;
    276 		if (copyin(iov->iov_base, cp, cnt))
    277 			break;
    278 
    279 		kth.ktr_len = cnt + sizeof(struct ktr_genio);
    280 
    281 		if (__predict_false(ktrwrite(l, &kth) != 0))
    282 			break;
    283 
    284 		iov->iov_base = (caddr_t)iov->iov_base + cnt;
    285 		iov->iov_len -= cnt;
    286 
    287 		if (iov->iov_len == 0)
    288 			iov++;
    289 
    290 		resid -= cnt;
    291 	}
    292 
    293 	free(ktp, M_TEMP);
    294 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    295 }
    296 
    297 void
    298 ktrpsig(l, sig, action, mask, code)
    299 	struct lwp *l;
    300 	int sig;
    301 	sig_t action;
    302 	sigset_t *mask;
    303 	int code;
    304 {
    305 	struct proc *p = l->l_proc;
    306 	struct ktr_header kth;
    307 	struct ktr_psig	kp;
    308 
    309 	p->p_traceflag |= KTRFAC_ACTIVE;
    310 	ktrinitheader(&kth, l, KTR_PSIG);
    311 	kp.signo = (char)sig;
    312 	kp.action = action;
    313 	kp.mask = *mask;
    314 	kp.code = code;
    315 	kth.ktr_buf = (caddr_t)&kp;
    316 	kth.ktr_len = sizeof(struct ktr_psig);
    317 
    318 	(void) ktrwrite(l, &kth);
    319 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    320 }
    321 
    322 void
    323 ktrcsw(l, out, user)
    324 	struct lwp *l;
    325 	int out;
    326 	int user;
    327 {
    328 	struct proc *p = l->l_proc;
    329 	struct ktr_header kth;
    330 	struct ktr_csw kc;
    331 
    332 	p->p_traceflag |= KTRFAC_ACTIVE;
    333 	ktrinitheader(&kth, l, KTR_CSW);
    334 	kc.out = out;
    335 	kc.user = user;
    336 	kth.ktr_buf = (caddr_t)&kc;
    337 	kth.ktr_len = sizeof(struct ktr_csw);
    338 
    339 	(void) ktrwrite(l, &kth);
    340 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    341 }
    342 
    343 void
    344 ktruser(l, id, addr, len, ustr)
    345 	struct lwp *l;
    346 	const char *id;
    347 	void *addr;
    348 	size_t len;
    349 	int ustr;
    350 {
    351 	struct proc *p = l->l_proc;
    352 	struct ktr_header kth;
    353 	struct ktr_user *ktp;
    354 	caddr_t user_dta;
    355 
    356 	p->p_traceflag |= KTRFAC_ACTIVE;
    357 	ktrinitheader(&kth, l, KTR_USER);
    358 	ktp = malloc(sizeof(struct ktr_user) + len, M_TEMP, M_WAITOK);
    359 	if (ustr) {
    360 		if (copyinstr(id, ktp->ktr_id, KTR_USER_MAXIDLEN, NULL) != 0)
    361 			ktp->ktr_id[0] = '\0';
    362 	} else
    363 		strncpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN);
    364 	ktp->ktr_id[KTR_USER_MAXIDLEN-1] = '\0';
    365 
    366 	user_dta = (caddr_t) ((char *)ktp + sizeof(struct ktr_user));
    367 	if (copyin(addr, (void *) user_dta, len) != 0)
    368 		len = 0;
    369 
    370 	kth.ktr_buf = (void *)ktp;
    371 	kth.ktr_len = sizeof(struct ktr_user) + len;
    372 	(void) ktrwrite(l, &kth);
    373 
    374 	free(ktp, M_TEMP);
    375 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    376 
    377 }
    378 
    379 void
    380 ktrmmsg(l, msgh, size)
    381 	struct lwp *l;
    382 	const void *msgh;
    383 	size_t size;
    384 {
    385 	struct proc *p = l->l_proc;
    386 	struct ktr_header kth;
    387 	struct ktr_mmsg	*kp;
    388 
    389 	p->p_traceflag |= KTRFAC_ACTIVE;
    390 	ktrinitheader(&kth, l, KTR_MMSG);
    391 
    392 	kp = (struct ktr_mmsg *)msgh;
    393 	kth.ktr_buf = (caddr_t)kp;
    394 	kth.ktr_len = size;
    395 	(void) ktrwrite(l, &kth);
    396 	p->p_traceflag &= ~KTRFAC_ACTIVE;
    397 }
    398 
    399 /* Interface and common routines */
    400 
    401 int
    402 ktrace_common(l, ops, facs, pid, fp)
    403 	struct lwp *l;
    404 	int ops;
    405 	int facs;
    406 	int pid;
    407 	struct file *fp;
    408 {
    409 	struct proc *curp = l->l_proc;
    410 	struct pgrp *pg;
    411 	struct proc *p;
    412 	int error = 0;
    413 	int ret = 0;
    414 	int one = 1;
    415 	int descend;
    416 
    417 	curp->p_traceflag |= KTRFAC_ACTIVE;
    418 	descend = ops & KTRFLAG_DESCEND;
    419 	facs = facs & ~((unsigned) KTRFAC_ROOT);
    420 
    421 	/*
    422 	 * Clear all uses of the tracefile
    423 	 */
    424 	if (KTROP(ops) == KTROP_CLEARFILE) {
    425 		proclist_lock_read();
    426 		for (p = LIST_FIRST(&allproc); p != NULL;
    427 		     p = LIST_NEXT(p, p_list)) {
    428 			if (ktrsamefile(p->p_tracep, fp)) {
    429 				if (ktrcanset(curp, p))
    430 					ktrderef(p);
    431 				else
    432 					error = EPERM;
    433 			}
    434 		}
    435 		proclist_unlock_read();
    436 		goto done;
    437 	}
    438 
    439 	/*
    440 	 * Mark fp non-blocking, to avoid problems from possible deadlocks.
    441 	 */
    442 
    443 	if (fp != NULL) {
    444 		fp->f_flag |= FNONBLOCK;
    445 		(*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&one, l);
    446 	}
    447 
    448 	/*
    449 	 * need something to (un)trace (XXX - why is this here?)
    450 	 */
    451 	if (!facs) {
    452 		error = EINVAL;
    453 		goto done;
    454 	}
    455 	/*
    456 	 * do it
    457 	 */
    458 	if (pid < 0) {
    459 		/*
    460 		 * by process group
    461 		 */
    462 		pg = pgfind(-pid);
    463 		if (pg == NULL) {
    464 			error = ESRCH;
    465 			goto done;
    466 		}
    467 		for (p = LIST_FIRST(&pg->pg_members); p != NULL;
    468 		     p = LIST_NEXT(p, p_pglist)) {
    469 			if (descend)
    470 				ret |= ktrsetchildren(l, p, ops, facs, fp);
    471 			else
    472 				ret |= ktrops(l, p, ops, facs, fp);
    473 		}
    474 
    475 	} else {
    476 		/*
    477 		 * by pid
    478 		 */
    479 		p = pfind(pid);
    480 		if (p == NULL) {
    481 			error = ESRCH;
    482 			goto done;
    483 		}
    484 		if (descend)
    485 			ret |= ktrsetchildren(l, p, ops, facs, fp);
    486 		else
    487 			ret |= ktrops(l, p, ops, facs, fp);
    488 	}
    489 	if (!ret)
    490 		error = EPERM;
    491 done:
    492 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
    493 	return (error);
    494 }
    495 
    496 /*
    497  * ktrace system call
    498  */
    499 /* ARGSUSED */
    500 int
    501 sys_fktrace(l, v, retval)
    502 	struct lwp *l;
    503 	void *v;
    504 	register_t *retval;
    505 {
    506 	struct sys_fktrace_args /* {
    507 		syscallarg(int) fd;
    508 		syscallarg(int) ops;
    509 		syscallarg(int) facs;
    510 		syscallarg(int) pid;
    511 	} */ *uap = v;
    512 	struct proc *curp;
    513 	struct file *fp = NULL;
    514 	struct filedesc *fdp = l->l_proc->p_fd;
    515 	int error;
    516 
    517 	curp = l->l_proc;
    518 	fdp = curp->p_fd;
    519 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
    520 		return (EBADF);
    521 
    522 	FILE_USE(fp);
    523 
    524 	if ((fp->f_flag & FWRITE) == 0)
    525 		error = EBADF;
    526 	else
    527 		error = ktrace_common(l, SCARG(uap, ops),
    528 		    SCARG(uap, facs), SCARG(uap, pid), fp);
    529 
    530 	FILE_UNUSE(fp, l);
    531 
    532 	return error;
    533 }
    534 
    535 /*
    536  * ktrace system call
    537  */
    538 /* ARGSUSED */
    539 int
    540 sys_ktrace(l, v, retval)
    541 	struct lwp *l;
    542 	void *v;
    543 	register_t *retval;
    544 {
    545 	struct sys_ktrace_args /* {
    546 		syscallarg(const char *) fname;
    547 		syscallarg(int) ops;
    548 		syscallarg(int) facs;
    549 		syscallarg(int) pid;
    550 	} */ *uap = v;
    551 	struct proc *curp = l->l_proc;
    552 	struct vnode *vp = NULL;
    553 	struct file *fp = NULL;
    554 	int ops = SCARG(uap, ops);
    555 	struct nameidata nd;
    556 	int error = 0;
    557 	int fd;
    558 
    559 	ops = KTROP(ops) | (ops & KTRFLAG_DESCEND);
    560 
    561 	curp->p_traceflag |= KTRFAC_ACTIVE;
    562 	if ((ops & KTROP_CLEAR) == 0) {
    563 		/*
    564 		 * an operation which requires a file argument.
    565 		 */
    566 		NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, fname),
    567 		    l);
    568 		if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) {
    569 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
    570 			return (error);
    571 		}
    572 		vp = nd.ni_vp;
    573 		VOP_UNLOCK(vp, 0);
    574 		if (vp->v_type != VREG) {
    575 			(void) vn_close(vp, FREAD|FWRITE, curp->p_ucred, l);
    576 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
    577 			return (EACCES);
    578 		}
    579 		/*
    580 		 * XXX This uses up a file descriptor slot in the
    581 		 * tracing process for the duration of this syscall.
    582 		 * This is not expected to be a problem.  If
    583 		 * falloc(NULL, ...) DTRT we could skip that part, but
    584 		 * that would require changing its interface to allow
    585 		 * the caller to pass in a ucred..
    586 		 *
    587 		 * This will FILE_USE the fp it returns, if any.
    588 		 * Keep it in use until we return.
    589 		 */
    590 		if ((error = falloc(curp, &fp, &fd)) != 0)
    591 			goto done;
    592 
    593 		fp->f_flag = FWRITE|FAPPEND;
    594 		fp->f_type = DTYPE_VNODE;
    595 		fp->f_ops = &vnops;
    596 		fp->f_data = (caddr_t)vp;
    597 		FILE_SET_MATURE(fp);
    598 		vp = NULL;
    599 	}
    600 	error = ktrace_common(l, SCARG(uap, ops), SCARG(uap, facs),
    601 	    SCARG(uap, pid), fp);
    602 done:
    603 	if (vp != NULL)
    604 		(void) vn_close(vp, FWRITE, curp->p_ucred, l);
    605 	if (fp != NULL) {
    606 		FILE_UNUSE(fp, l);	/* release file */
    607 		fdrelease(l, fd); 	/* release fd table slot */
    608 	}
    609 	return (error);
    610 }
    611 
    612 int
    613 ktrops(l, p, ops, facs, fp)
    614 	struct lwp *l;
    615 	struct proc *p;
    616 	int ops;
    617 	int facs;
    618 	struct file *fp;
    619 {
    620 	struct proc *curp = l->l_proc;
    621 
    622 	if (!ktrcanset(curp, p))
    623 		return (0);
    624 	if (KTROP(ops) == KTROP_SET) {
    625 		if (p->p_tracep != fp) {
    626 			/*
    627 			 * if trace file already in use, relinquish
    628 			 */
    629 			ktrderef(p);
    630 			p->p_tracep = fp;
    631 			ktradref(p);
    632 		}
    633 		p->p_traceflag |= facs;
    634 		if (curp->p_ucred->cr_uid == 0)
    635 			p->p_traceflag |= KTRFAC_ROOT;
    636 	} else {
    637 		/* KTROP_CLEAR */
    638 		if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
    639 			/* no more tracing */
    640 			ktrderef(p);
    641 		}
    642 	}
    643 
    644 	/*
    645 	 * Emit an emulation record, every time there is a ktrace
    646 	 * change/attach request.
    647 	 */
    648 	if (KTRPOINT(p, KTR_EMUL))
    649 		ktremul(l);
    650 #ifdef __HAVE_SYSCALL_INTERN
    651 	(*p->p_emul->e_syscall_intern)(p);
    652 #endif
    653 
    654 	return (1);
    655 }
    656 
    657 int
    658 ktrsetchildren(l, top, ops, facs, fp)
    659 	struct lwp *l;
    660 	struct proc *top;
    661 	int ops;
    662 	int facs;
    663 	struct file *fp;
    664 {
    665 	struct proc *p;
    666 	int ret = 0;
    667 
    668 	p = top;
    669 	for (;;) {
    670 		ret |= ktrops(l, p, ops, facs, fp);
    671 		/*
    672 		 * If this process has children, descend to them next,
    673 		 * otherwise do any siblings, and if done with this level,
    674 		 * follow back up the tree (but not past top).
    675 		 */
    676 		if (LIST_FIRST(&p->p_children) != NULL)
    677 			p = LIST_FIRST(&p->p_children);
    678 		else for (;;) {
    679 			if (p == top)
    680 				return (ret);
    681 			if (LIST_NEXT(p, p_sibling) != NULL) {
    682 				p = LIST_NEXT(p, p_sibling);
    683 				break;
    684 			}
    685 			p = p->p_pptr;
    686 		}
    687 	}
    688 	/*NOTREACHED*/
    689 }
    690 
    691 int
    692 ktrwrite(l, kth)
    693 	struct lwp *l;
    694 	struct ktr_header *kth;
    695 {
    696 	struct iovec aiov[2];
    697 	int error, tries;
    698 	struct uio auio;
    699 	struct file *fp;
    700 	struct proc *p;
    701 
    702 	p = l->l_proc;
    703 	fp = p->p_tracep;
    704 
    705 	if (fp == NULL)
    706 		return 0;
    707 
    708 	auio.uio_iov = &aiov[0];
    709 	auio.uio_offset = 0;
    710 	auio.uio_segflg = UIO_SYSSPACE;
    711 	auio.uio_rw = UIO_WRITE;
    712 	aiov[0].iov_base = (caddr_t)kth;
    713 	aiov[0].iov_len = sizeof(struct ktr_header);
    714 	auio.uio_resid = sizeof(struct ktr_header);
    715 	auio.uio_iovcnt = 1;
    716 	auio.uio_lwp = (struct lwp *)0;
    717 	if (kth->ktr_len > 0) {
    718 		auio.uio_iovcnt++;
    719 		aiov[1].iov_base = kth->ktr_buf;
    720 		aiov[1].iov_len = kth->ktr_len;
    721 		auio.uio_resid += kth->ktr_len;
    722 	}
    723 	kth->ktr_buf = (caddr_t)l->l_lid;
    724 
    725 	simple_lock(&fp->f_slock);
    726 	FILE_USE(fp);
    727 
    728 	tries = 0;
    729 	do {
    730 		error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio,
    731 		    fp->f_cred, FOF_UPDATE_OFFSET);
    732 		if (error == EWOULDBLOCK)
    733 		  	preempt(1);
    734 		else
    735 			tries++;
    736 	} while ((error == EWOULDBLOCK) && (tries < 3));
    737 	FILE_UNUSE(fp, NULL);
    738 
    739 	if (__predict_true(error == 0))
    740 		return (0);
    741 	/*
    742 	 * If error encountered, give up tracing on this vnode.  Don't report
    743 	 * EPIPE as this can easily happen with fktrace()/ktruss.
    744 	 */
    745 	if (error != EPIPE)
    746 		log(LOG_NOTICE,
    747 		    "ktrace write failed, errno %d, tracing stopped\n",
    748 		    error);
    749 	proclist_lock_read();
    750 	for (p = LIST_FIRST(&allproc); p != NULL; p = LIST_NEXT(p, p_list)) {
    751 		if (ktrsamefile(p->p_tracep, fp))
    752 			ktrderef(p);
    753 	}
    754 	proclist_unlock_read();
    755 
    756 	return (error);
    757 }
    758 
    759 /*
    760  * Return true if caller has permission to set the ktracing state
    761  * of target.  Essentially, the target can't possess any
    762  * more permissions than the caller.  KTRFAC_ROOT signifies that
    763  * root previously set the tracing status on the target process, and
    764  * so, only root may further change it.
    765  *
    766  * TODO: check groups.  use caller effective gid.
    767  */
    768 int
    769 ktrcanset(callp, targetp)
    770 	struct proc *callp;
    771 	struct proc *targetp;
    772 {
    773 	struct pcred *caller = callp->p_cred;
    774 	struct pcred *target = targetp->p_cred;
    775 
    776 	if ((caller->pc_ucred->cr_uid == target->p_ruid &&
    777 	     target->p_ruid == target->p_svuid &&
    778 	     caller->p_rgid == target->p_rgid &&	/* XXX */
    779 	     target->p_rgid == target->p_svgid &&
    780 	     (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
    781 	     (targetp->p_flag & P_SUGID) == 0) ||
    782 	     caller->pc_ucred->cr_uid == 0)
    783 		return (1);
    784 
    785 	return (0);
    786 }
    787 #endif /* KTRACE */
    788 
    789 /*
    790  * Put user defined entry to ktrace records.
    791  */
    792 int
    793 sys_utrace(l, v, retval)
    794 	struct lwp *l;
    795 	void *v;
    796 	register_t *retval;
    797 {
    798 #ifdef KTRACE
    799 	struct sys_utrace_args /* {
    800 		syscallarg(const char *) label;
    801 		syscallarg(void *) addr;
    802 		syscallarg(size_t) len;
    803 	} */ *uap = v;
    804 	struct proc *p = l->l_proc;
    805 	if (!KTRPOINT(p, KTR_USER))
    806 		return (0);
    807 
    808 	if (SCARG(uap, len) > KTR_USER_MAXLEN)
    809 		return (EINVAL);
    810 
    811 	ktruser(l, SCARG(uap, label), SCARG(uap, addr), SCARG(uap, len), 1);
    812 
    813 	return (0);
    814 #else /* !KTRACE */
    815 	return ENOSYS;
    816 #endif /* KTRACE */
    817 }
    818